摘要
视频全局运动(摄像机运动)所表现的视频序列之间的时间相关性,较其它视频特征更能表达视频序列的高层语义信息。为了能够有效快速的得到视频的全局运动,通过对视频分割技术的研究,提出了一种新的基于奇异值分解(SVD)的视频全局运动分割方法。该方法首先得到视频的切片图像,再利用奇异值分解(SVD)得到切片图像的方向图,然后统计切片方向图中每一行的角度直方图,最后根据摄像机不同运动类型对应的直方图模型,得到视频的运动分割结果。实验结果表明,所提出方法能够较准确的分割摄像机的运动,同时大幅度提高了分割速度。
Video's global motion (camera motion) behaves time-relativity of video sequence. It has more high semantic information than any other video's character. In order to acquire video's global motion more efficient, a novel method for video's global motion segmentation based on singular value decomposition (SVD) is presented. Firstly, video's slice image is obtained by slice operation. Secondly, slice orientation image's histogram is computed by using singular value decomposition (SVD), then, statistic each line's angle histogram of orientation image. Finally, video's global motion is obtained by cpmparing the angle histogram with the histogram model of camera motion. The experimental results show that the camera motion is segmented accurately, and the speed of segmentation is obviously improved.
出处
《计算机工程与设计》
CSCD
北大核心
2006年第23期4453-4456,共4页
Computer Engineering and Design
基金
河南省教育厅基金项目(sp200303099)
关键词
全局运动
运动分割
切片图像
奇异值分解
方向直方图
global motion
motion segmentation
slice image
singular value decomposition
directional histogram